In this episode of the Six Five Podcast Enterprising Insights, host Keith Kirkpatrick discusses the key highlights, areas of emphasis, and takeaways from Dreamforce, Salesforce‘s annual user conference. He discusses Agentforce, enhancements to Data Cloud, and the Salesforce Foundation program, among other announcements from the show, and closes out with his weekly Rant or Rave segment.
You can grab the video here and subscribe to our YouTube channel if you’ve not yet done so.
Listen to the audio here:
Disclaimer: The Enterprising Insights podcast is for information and entertainment purposes only. Over the course of this podcast, we may talk about companies that are publicly traded and we may even reference that fact and their equity share price, but please do not take anything that we say as a recommendation about what you should do with your investment dollars. We are not investment advisors and we do not ask that you treat us as such.
Transcript:
Keith Kirkpatrick: Hello, everyone. I’m Keith Kirkpatrick, Research Director with The Futurum Group, and I’d like to welcome you to Enterprising Insights. It’s our weekly podcast that explores the latest developments in the enterprise software market, and the technologies that underpin these platforms, applications and tools. This week I’d like to talk about Dreamforce, which is Salesforce’s annual user conference, covering all of the major announcements and key takeaways from the event. Then of course I’m going to get into my rant or rave segment where I take one element in the market and I will either champion it or criticize it. So let’s get right into it.
So Salesforce had their annual conference, it’s called Dreamforce, and once again, they held it in San Francisco, California. Lots of people there. It was great to see them generate those large crowds. I want to say it was like 40,000 people, perhaps a little more, all coming together around the Moscone Center in downtown San Francisco. And there were really a couple of major announcements, but really the main one that was sort of overriding everything was Salesforce’s full speed, total commitment to what they’re calling or their sales… I’m sorry, their Agentforce, which is really their AI driven agents. Now, what is this Agentforce? It sounds like something like the Air Force or something like that, but really what it is, it is the use of generative AI and other AI to create independent agents or technologies that are designed to conduct very, very specific tasks in order to ultimately help people do their jobs better and eventually complete tasks on their own.
So this Agentforce was a big deal because of the fact that if you think about what Salesforce has done over the past year with generative AI, they were talking in the past about this Einstein Copilot AI technology, which was really sort of designed around the idea of using generative AI to help people conduct tasks in a very general way. Well, what Agentforce is, it’s using this generative AI technology grounding all of that technology to very, very specific workflows, processes and data help within Data Cloud to help organizations conduct work more efficiently, help smooth out friction and processes, and ultimately take action independently of a human. Now, I want to be very clear that initially what we’re going to be seeing is again, a human in the loop, making sure that these things aren’t running amok. But I think the ultimate goal, of course, is to get to a point where this technology can handle low-level, repetitive tasks without human involvement or oversight. And obviously, the goal there is it allows humans to focus in on things that they do a much better job of handling. So things like connecting with customers, working through very, very complex situations, and making higher-level decisions that won’t necessarily be solely based on data that you could put into a data cloud.
So for example, if you think of a typical customer service interaction where there’s a product return. Generally speaking, that’s pretty straightforward, customer has a problem, they tell the company what’s wrong and hopefully the system has enough data that they could process that return, basically marking down, “Okay, here’s what was wrong with the item or what happened with the transaction.” And of course then a bunch of steps take place to process that return in terms of making sure that the customer has the information and the materials to send that product back. And then of course all of the other back-end things to make sure that all of that is accounted for within the company’s commerce system or ERP. Now, when we’re talking about doing things that are more complex, like perhaps having a return for a reason that is not covered under a traditional policy. And again, this example is not going to be great because ultimately the really, really challenging ones are the ones where it’s some crazy situation that almost never comes up, otherwise you would’ve a policy for it. So in that case, a human would be much better able to handle those types of situations.
So what we’re talking about here are processes that, generally speaking, tend to be fairly commonplace, at least on the service side or on the sales side, things like generating a quote or sending out a follow-up email. Things like that that are very, very well-defined. They are constrained by parameters that obviously companies will have to put into place, but ultimately the goal is to essentially take AI and deploy it in a very, very constrained way to address a job that needs to be done. So now Salesforce announced this week several pre-built agents that are actually available today. So for example, Service Agent, so this is what I was just speaking about, talking about handling service issues that are designed to improve customer service efficiency. So again, product returns or if someone has a question about a product or if there’s an issue with a service, routing them to the right department and then addressing their issue. So that would be Service Agent.
Sort of similar to that is something called Sales Development Representative. This is an agent that will engage with prospects anytime of day, anytime during the week 24/7, what have you. They will answer questions, manage objections, they can schedule meetings based on other inputs, based on data held within the CRM or even external data. And what this is really trying to do is again, address sort of the busy work or the administrative tasks that are very rote, that are very much straightforward, allowing human sellers to focus on what they do much better, which is building deeper customer relationships. Again, the focus is all on efficiency and productivity. A couple of other ones that they announced, Merchandiser, these are agents that are designed to help e-commerce merchandisers do again sort of very, very basic tasks, setting up their site, establishing goals, personalizing promotions, creating and editing product descriptions, identifying different insights and data, all of that type of activity that again, humans had done in the past.
But essentially, AI can be used to do it much more quickly, more efficiently, and let’s be honest, in some cases more accurately because they’ll be only looking at data and generally speaking, as long as the models have been pointed at the right information, which is technique that Salesforce has been talking about quite a bit called retrieval augmented generation, making sure those models are only grounded in vetted content or vetted information held within data cloud. As long as they have that, they generally will not screw up. And then of course, one of the other things that’s really important to mention is that these sorts of out-of-the-box agents are great, but the real power is of course when you start to customize them to work within a particular business. So in that case, they have something, Salesforce announced something called Agentforce Studio. Now this is a suite of low-code AI builders that allow organizations to customize existing agents or build new ones from scratch. So Agent Builder will use sort of existing tools like Flows, ProbTek, Templates, Apex and other APIs to help configure these agents using low code. Really what this does is it takes the need to deploy a developer to handle this and lets business users actually configure their agents by defining topics, providing specific instructions using natural language, and then creating a library of actions that the agent can choose from when the agent is doing its work.
Now, they also announced another couple of builders, a model builder, which is a low-code builder in control playing for registering testing and activating AI models and other LLMs of their choice across Salesforce. What does this mean? Well, the large LLMs, they do some things very well, but realistically they don’t do everything well. And in some cases, it’s like using a nuclear bomb to take out a clay pigeon. It’s just overkill, it’s too expensive, too much firepower. So what Model Builder does, it allows organizations to select the right AI model for the particular use case and industry. Now, Prompt Builder is another tool that was announced where it allows users to customize out-of-the-box prompt templates with their own CRM or data cloud data, and this will enhance the output of the generated results. So again, the idea here is you’re taking these prompts and you’re using information or using structures held within their CRM or data cloud to make sure that the prompt really only uses the data that you want it to as opposed to just sort of a generalized approach.
Now, what does this really mean? Well, the goal here for Salesforce is to make AI very, very easy to use and deploy. Their big mission throughout Dreamforce is obviously to publicize that they have agents, but also to really highlight the power that Salesforce has taking a platform-based approach saying, you as an organization don’t need to go out and build your own generative AI use cases and models on your own because really that’s not where… Companies often don’t realize the level of time, money, and effort it takes to do that and do that well. And another point of course is that it’s sort of like the old plumber analogy. If I wanted to replace all the pipes in my house, I could probably do it, but if something goes wrong, then it’s on me to fix it. And of course I might spend a ton of money and a ton of time doing it, and if something goes wrong, I may not know how and I might wind up calling in somebody else to help out. And by that point it would’ve just been cheaper to call in an expert plumber at the beginning. And that’s what Salesforce is saying is that they are building all of these tools and they have this platform that makes it easier and more efficient for an organization to try to deploy AI that way.
Now, if we’re going to take a step back, of course that’s what they want. That is their overall business strategy, which is to get folks onto that Salesforce platform to use because they are obviously, they’ve been moving toward more of a consumption-based model when it comes to Data Cloud. The more data that is actually imported or used by Data Cloud, they make more money. Same thing with their AI strategy. They are moving essentially, one of the things that they talked about this week, which is really interesting is they were talking about, I think it was called Salesforce Foundations. And what that is, it’s a program where they will, as Marc Benioff said, they are going to make it so if you’re enterprise level or above in terms of that being your customer level, you’ll get access to all of the different Salesforce. You’ll get access to Data Cloud and all relevant Salesforce clouds on a freemium basis, meaning you’ll have a limited amount of functionality. But the idea is that you’ll be able to use the technology, deploy agents, and see how all of that really works for your organization, being able to access data in any of those different applications and obviously having it all grounded into Data Cloud.
Now, there’s two things there. One is obviously it’s a stickiness play there. They obviously want organizations to start using all of these different applications because once you start using them, it’s that much harder to not use them. The second thing of course, is if we think about what they’re trying to do with pricing, again, they’re giving their customers a certain amount of users enough to get them hooked. It’s kind of like giving a kid just enough candy to get them to come back over and over and over again. And there’s nothing wrong with that approach. I think it’s actually quite good in that it does sort of feed into what Salesforce needs to do in terms of growing revenue. As AI becomes more commonplace, it will be used. More usage obviously equals more consumption of Salesforce and what they have to offer. Where they might run into, I don’t want to say a problem, but perhaps a challenge in the market is of course looking at other vendors out there that are taking it a step further and saying, “We’re not going to just charge you based on consumption.” Meaning, okay, you had X number of customer interactions, but we’re actually going to say we’re going to charge you based on outcomes. Meaning we’re not going to charge you if you can’t resolve if a customer is unable to resolve a particular problem, issue, complaint, what have you solely using an agent.
And the idea there is they are really saying, “We trust our technology so much that we are not going to just charge you on consumption. We’re going to actually look at the outcomes.” Now, realistically, we’re not going to know how this is all going to play out for quite some time until customers have had time to really get to know this approach, deploy it in the wild without any sort of pilot program or POC constraints there. I do think it’s also going to be very much dependent upon the type of company, the industry in which they operate, and all of that kind of stuff. It’s all going to play into it because for example, if we look at something like retail. Retail, generally speaking, you’re not going to have that many very, very difficult back and forth transactions or interactions, which by the way, the way Salesforce is looking at it is an interaction would not be just, it’s not like they’re trying to rack up charges going, okay, a customer contacted the agent and the agent wrote back and went back and forth 15 times. No to them, that’s one interaction. They’re not trying to nickel and dime customers, but the more that happens, that’s more compute that Salesforce is going to wind up accumulating costs for, as opposed to a simpler interaction where it might be one or two back and forths.
And where this becomes really interesting is if we look at different industry segments and different use cases. Something like healthcare, I could see that becoming the norm. Most people don’t interact with their healthcare providers for something simple at this point. Nobody wants to call them unless there’s a problem. And usually that problem is a little bit more complex and requires a lot of back and forth. Similarly, if you look at companies like telecommunication services providers, they typically have had pretty awful customer service scores because of the level of complexity in their offerings, particularly because they tend to not want to offer the same packages to every customer. They want to do it very personalized, but of course that makes it a much more complex situation when customers want to change packages, add features, delete features, and so on and so forth. So in that case, I think organizations are going to have to take a really hard look at how they might deploy agents and in which way in terms of how they consume and how they price or how they pay for it makes the most sense. And that’ll be on both sides from the vendor perspective, like Salesforce as well as the customer. So I think all of that’s going to play out.
But I think that the big takeaway here is that Salesforce is clearly leaning very, very heavily into this agent approach. I think Marc Benioff made some pretty bold predictions there in terms of the number of agents that will be deployed by the next Dreamforce. I don’t have the number off the top of my head, but it was a lot. And I think the challenge, again, is going to be demonstrating effectiveness of these agents, both in terms of if they do what they say they’re going to do, and he challenged everyone to go down and check out all the demos, which were great. And I sat in on a couple and the demos looked very good, but I would expect that. I can’t see why they would put a bunch of it that didn’t work well out on an exhibit floor. I think the bigger issue is how are they going to work in a real world environment, particularly when things aren’t neatly tied up. And what I mean by that is, most organizations do not just hold data within Data Cloud, they need to pull it in from other sources. Now that brings me to another sort of big point about Dreamforce.
Another big thing they kept talking about was their more open ecosystem approach. And what does that mean? Well, they are really kind of driving up the number of partners that they have through their Agentforce partner network. This is an open ecosystem that allows Agentforce agents to complete complex tasks by chaining together actions not only across Salesforce, but a whole network of other third party systems and agents. What does this mean? It means if you have data held in, AWS or in Google Cloud or Workday, Zoom, I think IBM and Box are the other partners that were announced as of this past week. It will be easy to get data from wherever it’s held within those services and then have Agentforce actually act upon that data. Why is that important? Well, as I was just starting to say, very few organizations operate in a heterogeneous stack environment, meaning they only use one platform, one vendor. Most have a number of different platforms, a number of different applications, whether due to the fact that they put things together piecemeal over time, which is the usual case or, and this is an important one, they want to make sure that they’re not locked into a sole vendor for both in terms of contracts, you don’t want to have all your eggs in one basket there. But also from a risk perspective, as we’ve seen with some unfortunate cyber hacking issues, you wouldn’t want to have all your data locked up. Most companies would not want to have all of their data locked up in a single platform just from a risk perspective.
So I think Salesforce is doing the right thing here by meeting customers where they are in terms of wherever they’re holding their data and allowing them to deploy these agents that work across the enterprise, as well as these other different applications and data stores. So let’s see here. Let me think if there’s anything else that really kind of jumped out at me. I think there were a couple of announcements there that also jumped out. One was with Data Cloud, they’re now supporting the ingestion of what they were calling unstructured data. That term is a little bit of a point of contention among some folks, meaning that some say, “Well, technically all data is structured.” It’s just, there are certain degrees of it. But I think what Salesforce is referring to is things like video, data held within video data held within chat data held within audio files, things where the data itself is not formatted in a neat sort of row and column format. That’s very easy to process, to capture and process.
So why is that important? Well, a lot of organizational information is held within those unstructured data formats and allowing Data Cloud to ingest that and process that information means that AI can work on it. And even more importantly, as we start to see things like, or applications like Slack become the place where workers go to not only collaborate on projects, just talking back and forth, but also pulling in other types of data, whether the data is held within Data Cloud or it’s held within Tableau, which is obviously another Salesforce application. When you have all that, you want to make sure that you’re able to access and utilize all types of organizational data. And as we see video and audio and some of these other formats become much more commonplace, it’s important that the AI and users are able to access that very easily. So I think that was another important announcement. And then I think the other thing, there are other announcements here within their different clouds. I think there were some announcements for AI enhancements for field service, certainly some for Marketing Cloud, for enabling more streamlined management of campaigns and use of marketing insights, all of that kind of stuff. And I have a research note that is going to be out in the next day or so if it’s not out already that I wrote in conjunction with my colleague Paul Nashawaty, where I go through all of the major announcements from Dreamforce. So I encourage you to take a look at that.
But I think the biggest takeaway here for me is that ultimately what Salesforce is really doing is trying to set up a platform that makes it easy for workers to access information from anywhere, activate that, whether it’s a human worker interacting with the data or an AI agent, or most likely a combination of the two, and being able to do that from whatever application they might be working in or prefer using. A lot of the talk is about using Slack as that sort of central place where you’d be able to pull in data visualizations from Tableau or pull in data from Data Cloud or information from Commerce Cloud, all of that. That’s all great. I think it really is reflective of the way that work will be done in the future, and particularly as we move from a very forms-based approach and moving into one where there’s going to be a lot more natural language-based querying of information using generative AI. Basically instead of pouring through a spreadsheet, you’d be able to just say, “Tell me what my sales were for the last week. What’s performing well, what’s not?” Interacting with the data in a much more natural way. The only way that really works efficiently is if the AI has access to all of that information wherever it is held. And ultimately, it all needs to be set up in a way that there is a single source of truth, meaning there can’t be millions of copies of the same data all over the place because you run into syncing issues and it just becomes impossible to manage.
So I think that is a great step in the right direction. Again, just to reiterate questions I have, obviously pricing is going to be an issue as they move forward. As always, this is something that was raised with me in a customer meeting. There’s the question of is generative AI something that any business will only want to work with one vendor? I don’t believe it is. I think that a lot of companies are going to, yes, they may have AI that they may base most of their activities around a platform, whether it’s Salesforce or ServiceNow or Oracle or you name the platform. There’s certainly room there, but I do think there is going to be a bit of risk mitigation by doing some homegrown development, using other vendors as well, because that is the way that you manage risk with AI. And then of course, the trend that also supports this view is most vendors are taking a very open approach and they’re trying to drive partnerships with other vendors to make sure that data can flow elegantly in both directions. So again, I think that Salesforce had a very good Dreamforce. It looked like it was very well attended. From what I heard, the entertainment was good. I didn’t actually get a chance to check out Pink at Oracle Park, but I suppose that everything went well there. So with that, I’m going to wrap this section up, and if anyone has any other feedback about Salesforce or Dreamforce, please do let me know in the comments.
Now I’d like to move on to my rant or rave segment. Now, this is where I take one item in the market and I will either criticize it or champion it. And this week I actually have a rant. Now, one of the things I’ve seen over the past couple of weeks, or actually more than a couple of weeks, it’s really been this whole season, has been when I go to conferences and I hear vendors or leadership from various vendors take the old approach of talking, not just talking about the benefits of their platform, but outwardly criticizing the performance of their competitors. Now, I know that’s a tactic as old as the world itself, but in a lot of ways I think it does the market and their customers a massive disservice. There’s two elements here. One is a lot of the stuff, it’s just blatantly commercial saying, “My AI can beat up your AI.” Well, of course you’re going to say that. There’s nothing new there. You’re not getting beyond what is really important, which is what is it that specifically that you do better than the other competition? And really it gets lost in there where the first emphasis is on the other company can’t do this, they’re having trouble doing this, they’re having trouble doing this. That seems to be the focal point when really it should be, let’s talk very specifically about what are the situations in which my solution will excel? That I think has much more value, much more relevancy, which really does resonate in the market.
Now, these companies will say, “Yes, we do talk about that with case studies and so forth”, but I think that when your CEO leads with a very, very bold statement in saying, “Only we can provide this type of stuff, this type of functionality with AI”, it automatically sets up the conference attendee to have a bit of skepticism in everything else that is said from there on. Now, the second reason why this is not necessarily a good strategy is there’s an old saying that you should never really punch down at your rivals. You should only punch up. And there is a little bit of punching down at some of the conferences I saw, which really doesn’t do you any favors because it almost puts you on that same level. Now, the other challenge is even if you are punching up, sometimes that can also backfire because if you really don’t have that advantage, it makes you look not credible. Now, I understand with all of this, it’s about messaging. It’s about trying to position the company in the best light in terms of functionality, but really what it’s all going to come down to is where are these customers or potential customers seeking to apply AI and then how? The message should be, how can my solution apply very much or very directly to your situation, your somewhat unique industry or use case, whatever that might be. And I think essentially, after you get past the high level of keynotes, we do get a little more of that. But I think the challenge is making that messaging stand out above the, “My AI can kick your AI’s butt” message, which I don’t think does anyone any favors. And in many cases it’s not necessarily provable anyway, so just more noise. And I don’t think it does anyone any favors, particularly buyers when they’re trying to make decisions in terms of their own strategy.
So with that, that was my rant for the week. And of course that’s all the time I have today. So I want to thank you all for joining me here on Enterprising Insights. I’ll be back again with another episode next week, focused in on the happenings within the enterprise application market. So be sure to subscribe, rate and review this podcast on your preferred platform, and I’ll see you next time.
Author Information
Keith has over 25 years of experience in research, marketing, and consulting-based fields.
He has authored in-depth reports and market forecast studies covering artificial intelligence, biometrics, data analytics, robotics, high performance computing, and quantum computing, with a specific focus on the use of these technologies within large enterprise organizations and SMBs. He has also established strong working relationships with the international technology vendor community and is a frequent speaker at industry conferences and events.
In his career as a financial and technology journalist he has written for national and trade publications, including BusinessWeek, CNBC.com, Investment Dealers’ Digest, The Red Herring, The Communications of the ACM, and Mobile Computing & Communications, among others.
He is a member of the Association of Independent Information Professionals (AIIP).
Keith holds dual Bachelor of Arts degrees in Magazine Journalism and Sociology from Syracuse University.
